𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴: As a Prompt Engineer, I contributed to fine-tuning and searching for the best parameters for Pulze’s Project Dream AI architectural visualization tool. This project had many challenges, as the model had to be fine-tuned from scratch, with huge image datasets. Which demanded many different techniques, including training a custom ControlNet, besides the more traditional techniques such as Dreambooth, Loras, and Negative Embeddings. In the process, I also had the opportunity to build custom tools for integration with A1111’s API.◦ 𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻: As the challenges and complexity of generative AI grow, so do the ways to mix and experiment with different models and solutions. Besides Stable Diffusion, I experimented with different LLMs and Multimodal Models for captioning and labeling datasets.

𝟮𝟬𝟮𝟯 𝗟𝗮𝘁𝗶𝗻 𝗚𝗥𝗔𝗠𝗠𝗬𝘀 𝗡𝗼𝗺𝗶𝗻𝗲𝗲 - 𝗕𝗲𝘀𝘁 𝗦𝗵𝗼𝗿𝘁 𝗙𝗼𝗿𝗺 𝗠𝘂𝘀𝗶𝗰 𝗩𝗶𝗱𝗲𝗼:◦ 𝗦𝘁𝗮𝗯𝗹𝗲 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻: I used Python and Stable Diffusion to move the music video’s character from the virtual to the real world playfully and believably. It was a very challenging project where I was able to learn several techniques for animation with AI and delve deeper into how Deep Learning works with diffusion models.◦ 𝗠𝗼𝗱𝗲𝗹 𝗙𝗶𝗻𝗲-𝘁𝘂𝗻𝗶𝗻𝗴: One of the main challenges of this work was to make a consistent animation that preserved the characteristics of the actress and the character’s costume. To reach this result, it was necessary to test several fine-tuning methods for Stable Diffusion, such as Dreambooth, Lora, and Textual Inversion.

* STORYTELLING: One of the main functions of a video editor is to transform raw material into a cohesive and meaningful whole. I was able to learn several storytelling techniques for the most varied types of products and target audiences.* TECHINICAL ASPECTS OF VIDEO: As a video editor and colorist, I learned many technical aspects of digital video such as pixel format, aspect ratio, frame ratio, color space, and chroma subsampling. This made my transition to computer vision easier.* DIGITAL MEDIA CONVERSION: During most of my work, I had to deal with conversions of different types of media files. This gave me a good understanding of codecs, compression algorithms, the trade-offs involving these technologies, and knowledge of hardware and operating systems. Also, I became familiar with tools like ffmpeg and shell.